Update README.md

README.md

@@ -766,22 +766,9 @@ configs:
 
 # WorldSpeech
 
-**53,635 hours of transcribed speech across 113 language–region configs**, drawn from national parliaments, public broadcasters, public-domain audiobooks, and international institutions (UN, ICC, Vatican Radio). Strong non-European and low-resource coverage across Sub-Saharan Africa, Asia, and Latin America.
+**53,635 hours of transcribed speech across 113 language–region configs**, drawn from national parliaments, public broadcasters, public-domain audiobooks, and international institutions (UN, ICC, Vatican Radio). 
 
-Every row consists of a 24 kHz speech utterance paired with a verified human transcript, an aligned ASR transcript, character error rate (CER) between the two, a WADA-SNR estimate, and four DNSMOS-P.835 quality scores — enough metadata to filter for TTS-grade, ASR-grade, or noise-robustness training without re-running quality estimation yourself.
-
-**Audio format:** OGG Opus, 24 kHz mono, ~64 kbps VBR
-**Splits:** 95 / 5 train / test (per config, seed 42)
-**Provenance fields:** `source`, `source_url`, `session_date`, `source_start_s`, `source_end_s`, `segment_id` on every row
-**License:** CC BY-NC 4.0, with per-config restrictions where the upstream license is more restrictive (see [Licenses](#licenses))
-
-> ⚠️ **Audio embedding caveat.** Five configs (`kh_km` Khmer, `la_lo` Lao, `mm_my` Burmese, `vn_vi` Vietnamese, `cn_ug` Uyghur — 3,941 hours total, all Radio Free Asia–sourced) include only metadata, transcripts, and source URLs. Audio must be downloaded separately from `source_url`. **Embedded-audio hours: ~49,694.** Total including metadata-only: 53,635.
-
----
-
-## Dataset License
-
-WorldSpeech is released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Individual source licenses apply per config and are listed in the [Licenses](#licenses) section. Where a source license is more restrictive than CC BY-NC 4.0, the more restrictive terms govern that config.
+Every row consists of a 24 kHz speech utterance paired with a verified human transcript, an aligned ASR transcript, character error rate (CER) between the two, a WADA-SNR estimate, and four DNSMOS-P.835 quality scores.
 
 ---
 
@@ -799,23 +786,7 @@ WorldSpeech is released under [CC BY-NC 4.0](https://creativecommons.org/license
 | Languages ≥ 50 h | 47 |
 | Weighted avg DNSMOS-P.835 OVR | 2.82 |
 | Sample rate | 24 kHz |
-| Codec | OGG Opus |
 
-### Coverage by region
-
-| Region | Hours | Configs |
-|---|---|---|
-| Europe |   23,364 |      26 |
-| East/SE Asia |   11,623 |      13 |
-| Latin America |    7,573 |      11 |
-| Sub-Saharan Africa |    3,493 |      29 |
-| South Asia |    2,098 |       9 |
-| MENA |    2,060 |      13 |
-| Pacific/Oceania |    1,095 |       5 |
-| Caucasus |    1,934 |       4 |
-| Central Asia |      394 |       3 |
-
----
 
 ## How to use
 
@@ -823,7 +794,6 @@ WorldSpeech is released under [CC BY-NC 4.0](https://creativecommons.org/license
 from datasets import load_dataset, Audio
 
 ds = load_dataset("AAdonis/WorldSpeech", "nl_nl", split="train")
-ds = ds.cast_column("audio", Audio(sampling_rate=24000))
 
 row = ds[0]
 wav    = row["audio"]["array"]         # 1-D float32 numpy array, 24 kHz mono
@@ -843,24 +813,6 @@ for row in ds:
     break
 ```
 
-### Filtering for TTS-grade subset
-
-```python
-ds_tts = ds.filter(lambda row:
-    row["dnsmos_ovr"] >= 3.2 and
-    row["snr"]        >= 15.0 and
-    row["cer"]        <= 0.05 and
-    1.0 <= row["duration"] <= 20.0
-)
-```
-
-| Use case | DNSMOS OVR | SNR (dB) | CER |
-|---|---|---|---|
-| TTS / voice cloning | ≥ 3.2 | ≥ 15 | ≤ 0.05 |
-| Clean ASR | ≥ 2.8 | ≥ 10 | ≤ 0.10 |
-| Robustness / noisy | — | — | ≤ 0.15 |
-
----
 
 ## Schema
 
@@ -1088,6 +1040,13 @@ Issues, takedowns, license questions: open a discussion on this repository.
 
 ---
 
+## Dataset License
+
+WorldSpeech is released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Individual source licenses apply per config and are listed in the [Licenses](#licenses) section. Where a source license is more restrictive than CC BY-NC 4.0, the more restrictive terms govern that config.
+
+> ⚠️ **Audio embedding caveat.** Five configs (`kh_km` Khmer, `la_lo` Lao, `mm_my` Burmese, `vn_vi` Vietnamese, `cn_ug` Uyghur — 3,941 hours total, all Radio Free Asia–sourced) include only metadata, transcripts, and source URLs. Audio must be downloaded separately from `source_url`. **Embedded-audio hours: ~49,694.** Total including metadata-only: 53,635.
+
+
 ## Licenses
 
 | Config | Country / Language | License | Notes |